Generative AI Islamic-Finance Product Adviser

The mandate

The board keeps asking where generative AI can improve explanation and service without being mistaken for authoritative product or Sharia judgement. Pilots draft answers from product documents, rulings, procedures and customer context, but management has not separated approved source interpretation, personalised assistance, financial recommendation and questions that must transfer to qualified human authorities.

The adviser will contribute three days monthly through a product-evidence review, a chair or supervisory-interface session, and preparation or individual conversations. Five Digital and Customer Committee meetings are included; an urgent release or incident question receives acknowledgement within one business day and a considered response within three, without the adviser issuing customer or Sharia determinations.

The term spans ten months and closes with the next AI product-and-control decision. The committee chair may propose that the board renew the appointment once for up to four months if a named supervisory or evidence question remains open; incomplete engineering delivery does not renew the term automatically.

This seat carries no line authority, board vote, product approval, customer decision or executive responsibility. Product, risk and compliance leaders retain operating accountability, and the Sharia Supervisory Council alone makes determinations within its remit; the adviser may clarify system boundaries and evidence but cannot endorse a product answer or ruling.

Other non-competing banking and technology work may continue after disclosure. A role with a direct digital-bank rival, material interest in a shortlisted model or knowledge vendor, contingent product economics, or confidential access to another institution’s supervisory interpretations creates a conflict requiring restricted papers, recusal or termination.

Why the board wants this voice

Directors understand banking and governance, while supervisory members understand the relevant principles, but nobody has deployed source-grounded generative products under both customer-conduct and specialist-authority constraints. Current proposals treat citations as proof that an answer stays within its permitted role. The board wants an operator who can expose where generation crosses from explanation into unowned judgement.

What you will own

• Press management to classify product knowledge, procedural guidance, personalised assistance, recommendation and supervisory interpretation as distinct AI use boundaries.
• Test source ingestion for authority, version, language, product applicability, effective date, contradiction and withdrawal before material enters retrieval.
• Challenge evaluations across Bahasa Indonesia, English, customer ambiguity, vulnerable circumstances, adversarial prompting and requests requiring qualified escalation.
• Shape response controls for citation, uncertainty, prohibited inference, identity, data minimisation, human transfer and preservation of the customer’s original question.
• Probe whether generated product comparisons or eligibility explanations create recommendation, disclosure or fairness consequences not captured by accuracy scores.
• Evaluate model and vendor options for private data use, source control, audit evidence, change notification, portability and termination.
• Frame the closing board record with approved use cells, prohibited judgements, accepted residual risk and evidence required before broader customer exposure.

Candidate qualifications

• Governed generative AI in regulated banking, insurance or payments where source fidelity and customer-conduct boundaries were material.
• Worked with Islamic-finance products and supervisory structures while respecting that technology leaders do not make Sharia determinations.
• Designed source-grounded knowledge systems with version, authority, contradiction, citation and withdrawal controls.
• Stopped or narrowed an AI customer journey after generated explanation crossed into recommendation or specialist judgement.
• Built multilingual evaluation that included customer ambiguity, escalation effectiveness and vulnerable circumstances beyond factual accuracy.
• Advised a board while managing conflicts among model vendors, product sponsors and independent control or supervisory functions.

Non-negotiables

• Can attend all five Jakarta committee meetings and both designated customer-operations or supervisory-interface sessions.
• Will disclose banking, model-provider, knowledge-platform and product investments before reviewing confidential papers.
• Accepts no authority to issue or represent a Sharia determination, product approval or individual customer recommendation.
• Brings deployed regulated generative-AI evidence; generic responsible-AI policy or banking governance alone is insufficient.

1. 49 words maximum. Describe one generated banking explanation you narrowed because it crossed into recommendation or specialist judgement.
2. 49 words maximum. Which current bank, vendor, product or investment relationship could constrain your independence here?
3. 49 words maximum. Confirm the Jakarta cadence and name the evidence required before a source-grounded answer reaches customers.